Re: [patch v3 3/3] percpu_counter: use atomic64 for counter in SMP

[Shaohua Li <shaohua.li@intel.com>]

On Fri, 2011-04-15 at 12:15 +0800, Tejun Heo wrote:
> Hello,
> 
> On Thu, Apr 14, 2011 at 10:04:50AM +0800, shaohua.li@intel.com wrote:
> > Index: linux/lib/percpu_counter.c
> > ===================================================================
> > --- linux.orig/lib/percpu_counter.c	2011-04-14 09:53:04.000000000 +0800
> > +++ linux/lib/percpu_counter.c	2011-04-14 10:01:29.000000000 +0800
> > @@ -59,13 +59,17 @@ void percpu_counter_set(struct percpu_co
> >  {
> >  	int cpu;
> >  
> > -	spin_lock(&fbc->lock);
> > +	/*
> > +	 * Don't really need to disable preempt here, just make sure there is
> > +	 * no big latency because of preemption
> > +	 */
> > +	preempt_disable();
> >  	for_each_possible_cpu(cpu) {
> >  		s32 *pcount = per_cpu_ptr(fbc->counters, cpu);
> >  		*pcount = 0;
> >  	}
> > -	fbc->count = amount;
> > -	spin_unlock(&fbc->lock);
> > +	atomic64_set(&fbc->count, amount);
> > +	preempt_enable();
> 
> Disabling preemption here doesn't make any sense.
> percpu_counter_set() inherently requires its users to guarantee that
> no other user is modifying the percpu counter.
ha, ok.
should I still rebase the patch against Christoph's patch? Looks that
one is still not settled down.